Auto-generated data sets for engineering panchromatic co-sensitized solar cells

Materials discovery can accelerated with data-centric approaches emerged from the synergy from experiments, simulations and theory. Artificial Intelligence (AI) could give support to find underlying patterns embedded in materials properties and assist reaches to understand the key variables that are more relevant in the design of new compounds. In this talk, I will present a rare example of a design-to-device approach to engineer efficient dye sensitized solar cells. This approach begins producing and analyzing auto-generated data-sets produced from mining data (text, tables and figures) in peer-reviewed literature and combined with high-throughput atomic scale simulations. As result of our study, we proposed five organic dyes, in which the combination of two of them produced in the laboratory an organic co-sensitized solar cell that showed an efficiency close to a high performance reference organometallic dye, the so-called N719. Additionally, I will present preliminary results of using AI approaches for inverse design, such as generative models, within our data sets to produce new dyes with targeted properties. Finally, I will discuss some of the opportunities that Argonne National Laboratory offers for students and early career scientists.